multiplayer

[wicked whale]

oh no, that's not what I meant. my RPC just sends a float or perhaps a vector, depending on how I do it. I'm just trying to reduce the amount of RPCs to send it

[thin stratus]

Then you'll need to limit it with your own timer, yes

[lethal wraith]

Hey guys. I was doing some code and I was wondering. Is it alright to call 2 different server RPCs inside a function?

[bright summit]

Yes you can

[thin stratus]

Sure but depends on why

[dry pebble]

thanks guys! Did you end up writing a tutorial on that or nah? I was never able to get to it

[thin stratus]

Not that I'm aware of. This whole thing is an general topic of GAS and CMC using two different (their own) approaches for Prediction.

The NetworkPredictionPlugin was a good candidate to solve that, but in the end is basically more on the side of the CMC in its approach, and still mostly incompatible with GAS in that regards.

Even if you put the Stamina into the CMC, you can't use GAS directly to alter it.
The only thing you can do is have a LOCAL ONLY Ability to start sprinting, or whatever consumes the Stamina.
And then have the CMC handle all the prediction etc. itself.

There was one solution to this on a Client's project, but it was rather complicated and I don't remember how that exactly worked.
Let alone feel good about sharing that exact solution. The idea behind that was to listen to GAS Adding and Removing Gameplay Effects, and housekeeping a local copy of whatever they caused in the CMC.

There was still the issue of Removal not being properly predictable fwiw.

[dry pebble]

ah thanks cedric, defs sounds best to keep them seperate still, thanks for sharing those details, there's a lot of nuance with multiplayer in UE sometimes

[limber gyro]

First time doing this kind of stuff, i have a game instance subsystem where i do my steam stuff, which in turn is gonna call my lobby player controller for the callbacks, does this code look OK? im worried my PC is not gonna exist and cause issues

[fiery wadi]

Hi, I am wondering how you would go about doing server functions within the controller of a pawn instead of doing it all inside a character blueprint ?

What I mean is that I have all the input mapping context inside the controller so the Input Enhanced Events and when I try to play an animation montage on all connected clients only the server see's the animation the clients do not.

I think the reference to the TopDown Charater is the main issue but im not sure what I would plug in here to call these events from the player controller and not inside the character itself. Thanks.

[limber gyro]

i have code for that in c++ gime a sec

u need to get the anim instance

[fiery wadi]

Ahhh ok thanks! 🙂

Hmm i cant seem to plug the anim instance into the montage node.

[haughty ingot]

There's a few things wrong here if your intention is to play an animation for all connections.

1.) You're only setting your `"As BP Top Down Character" on the server, "Event On Possess" does not get called from the controller's owning client.

2.) You're trying to call a multicast RPC on something that doesn't exist for anyone but the server and the owning client.

3.) You're calling another server RPC from a multicast RPC.

There's some basics you're not quite grasping here. I'd recommend reading up a bit more.
Here's a good starting point: https://cedric-neukirchen.net/docs/category/multiplayer-network-compendium/

[fiery wadi]

I figured that the controller would only exist on the server but just wondered if it was possible to do it this way instead of doing all the server calling within the character itself but will take another look at that article again. 🙂

[haughty ingot]

It exists on the server and the connection that owns it

[fiery wadi]

I,ll take a read of that article i find it quite hard to learn from documentation as it kinda takes a little while to "click" if you know what I mean sometimes you can read it but you cant see how you "apply" it in a real world scenario but i,m sure its a great article once i have abetter understanding. 🙂 so thank you 🙂

[limber gyro]

could someone with experience with steam/threads/subsystems look at this and tell me if this could some how go wrong pls?

[fiery wadi]

I find this article is quite interesting but it doesnt really show you a real world example or anything its more of a "This does this" more than this is how you can use "this" pretty much the same way a dictionary tells you the definition of a word.

"The Pawn/Character. They are possessed by the PlayerController and during that time, the PlayerController is the owner of the possessed Pawn. This means the Connection that owns this PlayerController also owns the Pawn."

But if I want to play a Montage from the Controller to it,s Controlled pawn (Or spedicially the mesh of tthe controlled pawn) I would imagine that you can use GetControlledPawn and then cast to the pawn's class and then get the mesh from that?

If the playercontroller owns that pawn then it should know about it right? or am i thinking wrongly ?

[dark parcel]

You are trying to multicast on player controller but remember that clients only know their player controller.
So even if you manage to play the montage on server, it will not play on other clients.

[fiery wadi]

Ahh so a multicast will never work on a player controller because it is not the server right?

[limber gyro]

rpcs have to follow a chain to get to where you want them to go

so if the PC only exists on the client and u call that rpc on the PC only that client will get the rpc

[dark parcel]

read the pinned material like suggested earlier

[limber gyro]

if u want montage animations to play on every character they should be called inside a character

[dark parcel]

12 times , practice, fail then repeat.

[limber gyro]

since those exist in every other client

[fiery wadi]

So how do i send an Input Enhanced Event from a Controller to a Pawn to then play a montage and spawn something ?

[limber gyro]

u just call the rpc chain on the character

rpc to server -> multicast to clients

[fiery wadi]

but the controller still needs a reference to the character to be able to call the RPC even inside the Charatcer yet ?

[limber gyro]

ur not gonna call the rpc inside the controller

[dark parcel]

you don't want to use multicast on anything stateful btw

don't pay heed to bp youtube multiplayer tutorial too much

mostly just straight up wrong and usses multicast for the wrong reason

If you want everyone to stay in sync and see the montage, then multicast is not the way

and you can read the pinned material why that's the case.

[fiery wadi]

Ok so I would call Input Mapping Context and stuff inside the character instead of the controller then so i can use the Enhanced Events inside the Character.

[dark parcel]

I can see that you are trying to do a casting animation. It's going to saves you a lot of time from reinventing the wheel to learn GAS. It comes with multiplayer flavour too

but just learn the basic replication from the pinned material for a start, it's the best source.

[fiery wadi]

I,ll try to understand that documentation at wizardcell also as I said though its kinda like reading a dictionary, I guess read and lots of trial and error is due 🙂

[dark parcel]

wizardcell and exi's compedium

[fiery wadi]

I think the main issue is that trying to understand what you need to do and where you need to do it. 🙂

Thank you for the advice. 🙂

[vapid gazelle]

Could use a sanity check. Am I correctly understanding that SceneComponent's bVisible is replicated?

    /** Whether to completely draw the primitive; if false, the primitive is not drawn, does not cast a shadow. */
    UPROPERTY(EditAnywhere, BlueprintReadOnly, ReplicatedUsing=OnRep_Visibility,  Category = Rendering, meta=(AllowPrivateAccess="true"))
    uint8 bVisible:1;

For some reason I had always assumed that visibility was not replicated and was to be set locally, but it looks like I might have been mistaken?

[haughty ingot]

It is

[dark parcel]

Did you manage to get it working? Setting the vissibility on server only does it for it self to me.

[haughty ingot]

I can’t remember off the top of my head but I believe I only applies to the root component

[dark parcel]

Ok nvm it is replicated

i didn't tick component replicate on the component 🤦‍♂️

[vapid gazelle]

Is the idea with bOwnerNoSee and bOnlyOwnerSee which are not replicated that the server will replicate the visibility to all of the clients (bVisible is replicated as we've established), and then locally you can fine-tune that visibility further by using the concept of network ownership?

E.g. maybe you want first person weapons to be visible, set on the server, but you want only the owner of the weapon to see the first person version of that primitive component?

[limber gyro]

i think its meant for first person weapons and meshes like you said, not sure if they had all that "refine visibility" in mind when they created those vars, i dont know tho but i would say they are just there because unreal was originaly an FPS engine?

[fiery wadi]

13.1 Multiplayer Golden RulePermalink
This is a super important rule, that diverges into three golden subrules of multiplayer, of which you should live by:

1 : Use replicated properties for replicating stateful events. (Rep Notify)
2 : Use Multicast/Client RPCs for replicating transient (not stateful) or cosmetic in nature events. (Multicast/Run on Owning Client) (Mainly for particle effects or nothing that impacts gameplay) ?
3 : Use Server RPCs to make the client communicate with the server (basically it’s the only way), and validate the passed data if needed. (Run On Server)

I think the bracketed things is the functionality that they are describing there. Just asking to ensure that I am thinking correctly about what t hey are saying.

[fossil spoke]

Those flags are local only, in that, if the Actor they are attached to has an outer chain that does/doesnt flow back to the local player, it will affect visibility.

They have nothing to do with replication of visibility.

The typical usecase is First Person

Where you might have a First Person Weapon as well as a Third Person Weapon.

Obviously you want to have fine grained control over which context sees what Weapon

[vapid gazelle]

Understood. Is sounds like for FPS games the combination of bVisible with bOnwerNoSee and bOnlyOnwerSee is most of what you'll typically need to manage weapon visibility, or is there something else that people often reach out for?

[fossil spoke]

bVisible and bHiddenInGame are pretty much the only options you have.

The owner visibility flags are less useful and more niche.

Ive pretty much only used them a handful of times outside of the FPS usecase above

[vapid gazelle]

Roger. bHiddenInGame is at the actor level, bVisible is for scenes within the Actor, if I recall correctly?

[limber gyro]

Guys im stuck on this since yesterday, my blueprint implementable events crash randomly

for some reason unreal cant find the functions

i moved all the steam related stuff to a subsystem yesterday

[dark parcel]

Look at the call stacks?

run on debug

[limber gyro]

call stack doesnt tell me anything rly

i know where the function is

and where it is being called

i rly have no idea where to look

[fossil spoke]

You should be passing weak pointers into the async task.

Also you still need to check the PC is valid in the Async Task before using it

[limber gyro]

do you think that may be causing the crash?

[fossil spoke]

Well if the PC is gone when the AsyncTask executes, then thats a problem isnt it

[limber gyro]

well but it shouldnt execute if it doesnt exist

am i missing something?

[fossil spoke]

You are changing threads

That doesnt happen instantly

TWeakObjectPtr<UPGAsyncAction_ListenForGameplayEventMessage> WeakThis(this);
            ListenerHandle = Router.RegisterListener(ChannelToRegister,
                [WeakThis](const FGameplayEventMessage& EventMessage)
                {
                    if (UPGAsyncAction_ListenForGameplayEventMessage* StrongThis = WeakThis.Get())
                    {
                        StrongThis->MessageReceived(EventMessage);
                    }
                });

Its much safer to use a Weak Ptr

[limber gyro]

so it can exist in a thread and not in another?

you can probably tell that i never coded with threads

[fossil spoke]

Also not sure why you are even passing this into the lambda

Your not using it

[lethal wraith]

Basically I am new to replication and I was doing an inventory system. I have a server rpc that adds the item to the player inventory and another server rpc that destroys the item that was picked up. Not sure if it is the best approach. I am also doubting that I should not use an RPC to update the inventory but i need to make some verifications in the server before adding the item to inventory, however I am not sure if replicating the inventory array without properly verifying if the item can be added to the inventory in the server can be problematic

[limber gyro]

some one helped me with the threads code

so i just went with it

ok so how do i make a weakpointer of the PC?

docs are saying i need to build them out of Shared References, Shared Pointers, or other Weak Pointers.

[fossil spoke]

TWeakObjectPtr<APlayerController> WeakPC(PlayerController);

In the same way I posted above

With this

[limber gyro]

IsValid doesnt like weakPointers

[fossil spoke]

You have to convert it into a strong pointer

if (UPGAsyncAction_ListenForGameplayEventMessage* StrongThis = WeakThis.Get())

So basically going back to a raw APlayerController*

[limber gyro]

first time in my life im hearing about strong pointers

so strong pointers are just the defautl ones?

[fossil spoke]

Not necessarily. Typically it just indicates that they will hold onto the resource.

[limber gyro]

i can do my first valid check on the regular pointer then? convert it, pass it, convert it back to strong and then check again?

[fossil spoke]

Weak meaning, it wont stop the resource from being removed

[fossil spoke]

Yes

That pattern is ideal when moving threads, or any latent actions.

Since the latent action might execute in the future where the resources dont exist anymore

So it would be unsafe to use them

[limber gyro]

does this look alright to you?

[fossil spoke]

Sure, you can use the StrongPointerPC inside the if instead of the WeakPC at that point

Id remove the this from the lambda capture to.

Just FYI, this is probably not going to fix the crash.

[limber gyro]

im gonna cry

[fossil spoke]

This is just an exercise in better practice.

It could fix it 🤷

[limber gyro]

can we do an exercice in fixing the crash? 😂

[fossil spoke]

But lets cross that bridge when we get to it

[limber gyro]

im gonna try this and see if it works

[thin stratus]

You usually only RPC things the server doesn't already know.

If you pick items up with an overlap for example, then no RPC is needed

If you pick them up with a button press, then you RPC the button press and do the rest on the server. Including determining what to pick up, modifying the inventory and destroying the actor

[lethal wraith]

I am using a button to pickup.

The problem is that when I add the whole logic into the server RPC it seems that it is not replicating the things correctly and I am not being able to destroy the item as the client and the item is not added correctly. The BP_Pickup implements an interact interface.

This is an image of my pickup bp, I get the data to spawn using an Id.

void UInventoryComponent::AddItemToInventory(const FName ItemId)
{
    auto ItemData = ItemsData->FindRow<FSItem>(ItemId, "");
    if (ItemData)
    {
        for (auto& item : InventoryItems)
        {
            if (item.Id == ItemId)
            {
                // Leave function if item is at max quantity
                if (ItemData->MaxQuantity == item.Quantity)
                {
                    GEngine->AddOnScreenDebugMessage(-1, 10.0f, FColor::Red,
                                FString::Printf(TEXT("Cannot add %s to inventory, inventory is full: %d"), *ItemData->Name, item.Quantity));
                    return;
                }

                // Add an item if there is already an item in the inventory
                item.Quantity++;
                GEngine->AddOnScreenDebugMessage(-1, 10.0f, FColor::Red,
            FString::Printf(TEXT("Added %s to inventory: %d"), *ItemData->Name, item.Quantity));
                return;
            }
        }

        // Create the item in the inventory if it does not exist
        FInventoryItem InventoryItem;
        InventoryItem.Id = ItemId;
        InventoryItem.Quantity = 1;
        InventoryItem.MaxQuantity = ItemData->MaxQuantity;
        InventoryItems.Add(InventoryItem);

        GEngine->AddOnScreenDebugMessage(-1, 10.0f, FColor::Red,
            FString::Printf(TEXT("Added %s to inventory: %d"), *ItemData->Name, InventoryItem.Quantity));
    }
}```

[lethal wraith]

I can just execute the function in the server but I do not get the replicated version in my client. I could add another client rpc to get the replicated version but I think it should be avoided when possible as far as I know (btw this is the continuation of the above message)

[thin stratus]

The inventory variable, usually an array, would be marked as replicated

Which removes the need for the client RPC

[limber gyro]

@fossil spoke i tried a bunch of times and its not crashing

[fossil spoke]

Thats good.

[limber gyro]

im still not 100% sure

thanks for the help tho

hopefully its fixed and im just paranoid

[lethal wraith]

It is, somehow it is not replicating. I created a function that prints the inventory of every client in the server and it is not being replicated

[fossil spoke]

As I said, that might not necessarily have been the issue.

But its what stands out given the code you were originally using.

[lethal wraith]

Also not being able to destroy the actor :/ I guess it has to do with the fact that my player does not have ownership over the item?

[fossil spoke]

When working with Lambdas and any latent/async actions, its much better/safer practice to pass WeakPtrs into them

[limber gyro]

i will keep that in mind going forward, i dont know much about multi threading, if it starts crashing i will bother people here again. Hopefully its fixed and i dont have to deal with steam ever again

[fossil spoke]

Im pretty sure i recommended the RedpointEOS plugin to you earlier?

You will have a much easier time doing all of this if you use that instead.

Just my opinion.

[limber gyro]

you did, but im using .26 and between not knowing if it is even compatible and re doing the whole social stuff i decided to try to fix it

i even bookmarked it for future projects

[fossil spoke]

Yeah, you would be unable to get it properly working for 26

Its way out of support now for that version

[vapid gazelle]

Is an Actor Reset() function typically only called on on the server, or do I simply have things misconfigured on my end? I'm calling Reset Level from my Game Mode, which runs only on server, but I'm wondering if I should be expecting that Reset to be replicated to clients as well somehow? Or is this left up to the programmer to decide?

From looking at the source of GameModeBase, it doesn't look like it cares much for whether the Actor is replicated from the server, so I'm guessing that's left up to the implementer. Hoping to sanity check with those who might know.

void AGameModeBase::ResetLevel()
{
    UE_LOG(LogGameMode, Verbose, TEXT("Reset %s"), *GetName());

    // Reset ALL controllers first
    for (FConstControllerIterator Iterator = GetWorld()->GetControllerIterator(); Iterator; ++Iterator)
    {
        AController* Controller = Iterator->Get();
        APlayerController* PlayerController = Cast<APlayerController>(Controller);
        if (PlayerController)
        {
            PlayerController->ClientReset();
        }
        Controller->Reset();
    }

    // Reset all actors (except controllers, the GameMode, and any other actors specified by ShouldReset())
    for (FActorIterator It(GetWorld()); It; ++It)
    {
        AActor* A = *It;
        if (IsValid(A) && A != this && !A->IsA<AController>() && ShouldReset(A))
        {
            A->Reset();
        }
    }

    // Reset the GameMode
    Reset();

    // Notify the level script that the level has been reset
    ALevelScriptActor* LevelScript = GetWorld()->GetLevelScriptActor();
    if (LevelScript)
    {
        LevelScript->LevelReset();
    }
}

[fossil spoke]

Reset is only called on the Server, that is correct.

[vapid gazelle]

Say I have an Actor with a Collision component, both set to replicate. Do the collision settings of said Collision component replicate from the server to the client, or are to be set locally by each instance of the game? Unfortunately not super obvious just by looking at the source given that it's not as simple as a couple of properties to check, from what I can tell.

My impression is that the answer is no, collision settings are to be configured locally to each instance, but I would love to sanity check.

[fossil spoke]

Collision settings are not replicated.

[vapid gazelle]

Thank you.

[violet walrus]

In a Multiplayer co-op game where you have a save file, what would you do to reload from that save file?
Assuming there is clients connected already and you are already in gameplay

In singleplayer you would simply re-open same level and let it load like it's first time

[dark parcel]

Thats entirely up to your design won't it.

Some games only allow the host to save. Others everyone keep a copy

I guess we are talking about listen servers only.

I'm not sure if everyone keeps a copy in baldurs gate

[violet walrus]

Yeah, only host has a save. My question was "how would you handle it when host tries to load from that save"

[dark parcel]

Or just the host

[violet walrus]

Like "reload previous checkpoint"

[dark parcel]

The save file should contain all the data you need to reproduce the same state.

[violet walrus]

Yeah, minus the part after you saved

But you can't simply re-open same level in MP, can you?

What happens to clients?

[dark parcel]

There are hard travel and server travel.

Forgot how they work again.

But with server travel no one gets disconnected.

Don't quote me on hard travel (open level) but I think everyone gets disconnected then reconnected 🤷‍♂️

https://dev.epicgames.com/documentation/en-us/unreal-engine/travelling-in-multiplayer-in-unreal-engine

I think you hard travel once when you host and opened a level. And when client joins for the first time.

afterward you should use server travel

For the server only.

Will jump the server to a new world/level.

All connected clients will follow.

This is the way multiplayer games travel from map to map, and the server is the one in charge to call this function.

The server will call APlayerController::ClientTravel for all client players that are connected.

I would guess that it would be like re-opening the same level if you jump to the same level.

[fossil spoke]

@dark parcel @violet walrus A hard travel causes Clients to disconnect. A Seamless travel does not.

Seamless travels can only occur when the GameMode has that feature enabled (its enabled by default) and they only occur for the Server you are connected to.

For example

A hard travel always occurs when you first connect to a Server

When that Server switches to a new level, it will perform a Seamless travel for all Clients.

If a Client leaves that Server, that Client hard travels again to whereever they are going.

[dark parcel]

Awesome, so I shouldn't over think it.

Hosting new game = Hard travel for server
Joining a game = Hard travel for client

from there on just server travel

I guess?

[fossil spoke]

Calling ServerTravel just causes the Server to switch levels, it has nothing to do with a hard or seamless travel.

[violet walrus]

So can I just reload same level to reset state just like in Singleplayer?

(Using ServerTravel)

[fossil spoke]

From memory, that was never a good solution.

Typically you should just call Reset

For example, we use AGameModeBase::ResetLevel to do match "rounds"

Like in CoD or CS

[violet walrus]

So can it reset the world just like re-opening same map?

[fossil spoke]

Every Actor has Reset called on it

[violet walrus]

Never used it

[fossil spoke]

Its a function that the GameMode calls

On all Actors

So you can determine what state that Actor should return to when the level is reset.

If you want a total clean slate, then yeah, maybe reopening the level is better

But as I said earlier, from memory, that had issues

[violet walrus]

So it's manual implementation

[fossil spoke]

But those issues might be fixed

[fossil spoke]

Using ResetLevel gives you more control, as you can persist state much easier.

[violet walrus]

Not like calling AGameModeBase::ResetLevel does a lot. It just same thing as I'd implemted similar interface and called reset on each actor

[fossil spoke]

Instead of doing a full level transition

It is very manual yes.

[pseudo wagon]

On mobile, does game receive reliable RPC when app is background?

[fossil spoke]

It should eventually.

Probably better to ask in #mobile

Since im not sure what state the App is in when its in background mode for Mobile.

[violet walrus]

Is it server only or replicated?

[fossil spoke]

Its only called by the Server.

[pseudo wagon]

From my testing, it's seem not.

[fossil spoke]

Then there is more to it that I dont know. I have not developed for Mobile in UE

Good luck 😦

lol

[dark parcel]

[violet walrus]

UE + Mobile + MP

https://tenor.com/view/rainbow-pride-gif-26074061

[fossil spoke]

Yeah pretty much

Thats like the worst use case

Maybe tack on 2D as well

Just to be sure

lol

[dark parcel]

I feel bad for people doing 2D in unreal and pixel streaming

at least epic nuked the web support

while the other just abbandoned.

[pseudo wagon]

Oh no, the project I am working on is multiplayer rouge-like mobile game.

[dark parcel]

Solo dev?

how does match making even work for mobile

[pseudo wagon]

You can call it is an indie company. 🤣

[dark parcel]

As long you have someone to share the pain, it can be fun

[pseudo wagon]

Nah, my team mate prefer use dirty instead of fix from root cause, that make me in pain.

[dark parcel]

😔

[unkempt tiger]

Is there some flag I can check to see if an actor fully got its first net bunch?

Oh I guess I can check if HasActorBegunPlay according to this function 🤔

[odd hamlet]

Did anyone encounter this error?
it appears at this part of the code

void AThirdPersonCharacter::GetLifetimeReplicatedProps(TArray<FLifetimeProperty>& OutLifetimeProps) const
{
    Super::GetLifetimeReplicatedProps(OutLifetimeProps);

    DOREPLIFETIME(AThirdPersonCharacter, PlayerLobbyInfo);

}```
And it's strange, because sometimes it runs as it should, and sometimes not

[karmic briar]

Rep notify?

[nova wasp]

you are gauranteeing that a field called PlayerLobbyInfo exists in code, if this fails you probably messed up the declaration of PlayerLobbyInfo

you have not shown that

[odd hamlet]

yep

[odd hamlet]

and this is my PlayerLobbyInfo

[nova wasp]

hot reloaded dll

never use hot reload after making header changes

you must build and run from your IDE if you don't know what you are doing to not corrupt things like this... check the pins in #cpp for more details

afaik patch_n implies hot reloading but I could be wrong...

[odd hamlet]

I am prettty sure i ain`t using hot reload

[odd hamlet]

just using this button

[nova wasp]

do not use that button to compile any header changes

it will break your project in subtle ways

[odd hamlet]

oh so how am I suppose to compile then ?:))

[pseudo wagon]

Turn UE off and on.

[odd hamlet]

wtf?

close it and open again?

[nova wasp]

run + build from an IDE

[odd hamlet]

like close, clean and rebuild, and then open again, and it`s like it compiles itself automatically?

[nova wasp]

no, clicking on the project is also wrong as that will use the latest development editor binary by default without changing it

[odd hamlet]

what clicking on the project bro

[dark parcel]

@odd hamlet https://landelare.github.io/2023/01/07/cpp-speedrun.html

[stuck hare]

Hi, I'm new at working with multiplayer
I've made a project with client/server stuff, everything works fine in editor

I have no clue where to go from there, and how to make a working executable with my project that will connect players to the server

Are there any ressource someone could recommand that explains how to do this?

[dark parcel]

decide on the server first?

are you gonna have players hosting the server them self (listen server)

or you want to deploy your own dedicated servers

[stuck hare]

I'd like the least complicated option to setup
From what I could gather that seems to be listen server?

[dark parcel]

well dedicated gonna cost you money.

[stuck hare]

yeah no this is just a super simple sample project I'm doing in my free time

[dark parcel]

You can use steam, I reckon that would be the most viable option

testing is free, but you will have to cough $100 when you want to publish the game.

[exotic wasp]

dedicated servers don't have to cost money. they can still be hosted by someone normally from their own machine

[dark parcel]

true

[stuck hare]

I don't plan on publishing, but I do want to be able to send this to someone else and have them test it on their side

[exotic wasp]

listen is by far the most convenient for players

[dark parcel]

you can use steam and use their test app

[stuck hare]

Cool, is there a specific tutorial you would recommend?

[dark parcel]

there are tutorials out there

you just have to pick one and hope it work I guess

[lament flax]

if you do a system where a player can push slowly a cube using addforce on tick (using player inputs as direction), how would you do so the player CMC speed is always not to high/low so it can stay close enough to the cube ?

for now im moving my cube with AddForce while simulating physics (because the object can fall, go up, etc)
since its MP it means that multiple players can "grab" the cube, and as said all the grabbing player inputs are used to move the cube. The more the players going forward in the same direction, the faster will the cube move

the hard part that i still didnt yet tackle is, how to make the pushed cube not increase in speed (because for reason, the velocity increases even with constant AddForce)
for the CMC i wonder if not using it while moving but just use a SET LOCATION would work, but i guess ill miss all the networkd correction/prediction part

[thin stratus]

@lament flax Idk, that whole idea sounds flawed. Physics in Multiplayer. Multiple players interacting with it.

This is correction heaven

[lament flax]

what would be a cheap solution ?

or, without considering heavy lag, is the idea still flawed ?

does physics have no correction ?

[vapid gazelle]

In the Network Emulation settings of the editor, what exactly does the Emulation Target enum do as far as introducing latency? I'm not fully grokking what it means to apply it to server only, client only, or both.

UENUM()
enum class NetworkEmulationTarget
{
    Server UMETA(DisplayName = "Server Only"),
    Client UMETA(DisplayName = "Clients Only"),
    Any UMETA(DisplayName = "Everyone"),
};

[lament flax]

also, is there a clean way to override CMC velocity ?

[crisp shard]

i've never tried to make something where you would change pawns, but im curious on how much of a task it is to implement something like this ? would i need to change controllers or would it just be pawns and would you lose any data / variables from changing from one to the other ?

for example, something as simple as controlling a vehicle or turret like weaponry or soemthing, not sure how crazy that would be to change the pawn, or maybe i wouldn't even need to do so, maybe just do some weird stuff in the character

[dark edge]

just possess it

say it was GTA, you could do "get in the car and drive it" as:
Play animation
Attach character to car pawn
Possess car
Now you're driving the car

[crisp shard]

oh very nice, simpler than i imagined. so i could still deal dmg and what not to character or any char related events via that controlled pawn

[dark edge]

assuming your system doesn't depend on the character being currently possessed by a player, sure

[crisp shard]

nice, and let's say you're controlling something that can go far away from the originally controlled character, would relevancy / anything like this be something i should be aware of ? for example, the character gets killed while controlling the pawn , i.e. do the events on that character still fire on the client that is controlling the other pawn ? and this way if you die, you'd lose control of the pawn and i could destroy that or whatever the logci would be

[dark edge]

Depends. You'd probably want some sort of persistent link to character that you wouldn't necessarily want for a car

if a car you drove got destroyed on the other side of the city, who cares

[crisp shard]

right, and in the scenario where the character is not being controlled ( the character is controlling another pawn) but still in world, and i go far away from my character relatively, if the character were to get attacked and killed bcasue it would just be sitting there and not controlling itself, i'd just want to make sure that client's character doesn't die and they are still controlling some other pawn

[dark edge]

I mean it depends

if it's Cyberpunk you'd die

[crisp shard]

yea for sure, in my case i just want the character to stay vulnerable as if they werent controlling the pawn so that if they don't pay attention i suupose they will die lol , and thus losing control of the other pawn

[dark edge]

Yeah it could be as simple as having a char ref somewhere MyCharacter which is valid even when not possessed, you could then be bound to an OnDeath or OnDamage dispatcher on it

OnDeath -> unpossess MyCurrentPawn -> show death screen

that would work for the case where you're piloting your meat suit and for when you're elsewhere

[crisp shard]

very dope indeed, i like it. going to give some things a try and report back in a few days w findings

[vapid gazelle]

Could use some advice for this bizarre issue.

How do I track down what is spawning a player character in my codebase? When I turn on network emulation and joinmy game as a client, then once in a while I end up with an extra ACharacter being spawned at 0,0,0 in my map with no indication of what actually spawned it, how it ended up there. I put breakpoints on every single Spawn Actor call I could find and it doesn't appear that the extra character is created through any of them, only what I expect to be spawned spawn there.

It appears to be the Default Pawn Class of my Game Mode, when I change it, that's the class of the actor that gets dropped in the middle of the map once every few restarts.

Apparently if I set my Default Pawn Class to None this doesn't reproduce anymore, but I wish I had a better understanding of why that happens. Maybe some race condition around the game mode spawning the player character for my controller to possess by default and then not having the time to clean up when I don't end up using it?

[quasi tide]

GameMode spawns the player. HandleStartingNewPlayer or something like that.

[vapid gazelle]

Yeah, oddly enough I have Start Players as Spectators turned on, which is why I think this doesn't reproduce consistently.

[thin stratus]

@vapid gazelle Not really much to suggest here, other than what you already figured out.

If you are manually spawning your Characters, then you usually set the DefaultPawn to None, cause otherwise, each Client (ListenServer included) gets a Pawn spawned by that.

In a lot of cases, spawning the Character by hand isn't needed though as you have all sorts of overrides in the GameMode to define the actual DefaultPawnClass dynamically, and also figure out where to spawn the Character/Pawn.

Your best guess will be breakpointing the SpawnDefaultPawnForPlayer functions (idk if that's the exact name) and see when they are called and when not.

[vapid gazelle]

That's very helpful, thank you 🙏

[lament flax]

for some reasons i thought UE had some built in correction for physics

so i didnt thought it would be an issue

[thin stratus]

Na, quite the opposite

All of your Players have predicted Movement, which will move the Block locally first and then tell the Server to do the same, which then gets send to the Clients.

Which means every Player will think they are right, and none will be.

[limber gyro]

Is "initNewPlayer" called after a recconect or just the first time a player connects?

also when is "CopyProperties" called?

im trying to figure out the proper way to set data on disconnects and reconnects

[thin stratus]

Best way to answer that is to go to the function and look for its usages in the code-base

I can't presicely answer that without doing the exact same now, so I think it's probably better if you do it yourself to learn

[vapid gazelle]

1. Generally speaking, how many times can I expect an OnRep function to be called on the client for a particular replicated property, assuming there's some latency and packet loss involved? I'm noticing that some of the time the OnRep gets called more than once, even though there was only one update on the server to that particular property. Even more times depending on how poor the emulated network conditions.

I'm assuming there's no theoretical upper limit to this, and to be resilient clients should expect to have OnRep fire many times

2. Am I correctly understanding that the server will not trigger replication if a replicated property is never modified at runtime? I suppose there is no need for replication in that case because the constructor will have made the same exact copy on the client and the server?

3. I have this corner case where I'm repping a TList which points at Actors that are replicated to the client. Sometimes the list is replicated before the Actor has had a chance to be. It seems like the engine will re-play the OnRep function once the Actor pointed at by the TList finally replicates. Am I just hallucinating this?

[thin stratus]

1. In theory only once per change, unless the packages arrive in the wrong order, which I'm not sure would actually matter and have that effect.
   The only other case is if you actively set it to Notify Always and you are constantly setting it to the same value somewhere.

2. If the value of the variable is set by the Server on spawn of the Actor, then you gotta mark it as replicated. Otherwise the Client will load the Actor with the default value.

3. Yes, cause the List or Array will replicate with its changed size, but with a nullptr in it, due to the client not being able to resolve the instance the NetGUID is supposed to point to yet. Once the Actor gets replicated, the OnRep should call again, now with the correct pointer.

One could argue that point 3. means that for 1. there is another case, but ultimately the OnRep for the Array calls for different reasons and not the same value twice.

I think the same can happen with a normal pointer, assuming it is pointing to ActorA, if the Server spawns ActorB and sets the pointer to it, it might first OnRep with a nullptr and then with the actual actor pointer.
Generally suggested to test your pointers for being valid inside OnReps

Also keep in mind that OnReps might call before BeginPlay for connecting players, which might cause changes in the OnRep to not fully work if they are guarded if HasBegunPlay.

In that cause you'd need to call the OnRep manually on BeginPlay (Client only fwiw).

That's an edge case fwiw

[stuck hare]

Hi, working on a multiplayer card game

I finished making the game, and everything worked fine when testing in editor with two clients
I'm now trying to implement the Advanced Steam Sessions plugin

I've managed to make it work somewhat, but now one of my client is not a client anymore, and seems to only be showing stuff that's on the server (it doesn't show some of the info that's client-side only, client functions don't work, etc)

I'm very confused atm, can someone explain to me what I'm missing?

[vapid gazelle]

Really useful context, thanks a ton. Saving that for my records.

[vapid gazelle]

It almost feels like there should be a checklist of these rules to go through any time you make a new Actor with replication inside. It's so easy to forget to bulletproof the class against some of these race conditions and then spend hours hunting for them days or weeks later when someone's packet sequence gets sufficiently unlucky and they repro the corner case.

[thin stratus]

Yeah, knowing all of that and taking care of it is basically the next level :P

You never stop learning after all

[limber gyro]

In case any 1 wants to know "overrideWith()" is called on postLogin if i didnt read it wrong.

so data is re added AFTER "ChoosePlayerStart()", so if any 1 has issues with the player not going to the proper player start on reconnect u might need to call "ChoosePlayerStart()" again manually, thats what i am going for anyways

[limber gyro]

When players with not so good connectinons host a session characters become jittery, is there a var that i can change to help with this?

[fossil spoke]

Not really. You cant make a bad connection better by magic.

[limber gyro]

well but there has to be something that would make those corrections less noticeable no?

like a correction distance or something

altough im calling them corrections they are not corrections i think

its just the regular interpolation

at least i am assuming it is

[fossil spoke]

Corrections are based on updates incoming from the Server, if the Server has really choppy updates there isnt anything you can change to fix that.

[limber gyro]

what if the client is bad and is sending choppy data, is there anything that can be done?

overwatch for example, every enemy character and team mate feels smooth and i dont belive every player i encountered in that game has a good connection so there has to be something that can be done

is it a unreal limitation?

[dark edge]

players aren't hosting Overwatch

dedicated servers are

[limber gyro]

so dedicated servers make that much of a diference?

[quasi tide]

They can

Also, their prediction/rollback could be implemented in a different way (it is, after all, it is ECS) and be more lenient/strict

[graceful flame]

So for example if a client is running the game at only 20fps but the server has a tick rate of 60, then other players will see the player with low FPS still having smooth animations but they'll appear to be standing around or react slower than those at a higher framerate because inputs sent from the client to the server won't be received as often as they could be.

Is that correct?

[nova wasp]

It depends on what exactly is being sent from that client to the server... there are a few things that will definitely get horrible at low client fps

[graceful flame]

Makes sense, I guess that's why prediction methods such as dead reckoning exist for like driving games or boat games. Just use whatever the last input the client sent and presume it remains the same until some new input is received aka keep going in a straight line or keep making a left turn etc.

[nova wasp]

this is also called extrapolation if that helps... the default CMC setup will actually do exactly this more or less

a lot of different things come into play here

for example higher acceleration will mean differences and mispredictions will have larger differences from the current

and you can sacrifice recency for smoothness a bit

lots of different tweakable things can change how cmc smoothing works

[limber gyro]

what can i tweak exactly?

[nova wasp]

a lot... 😅

[limber gyro]

out of the box that is

[nova wasp]

To start with you can change the math sim proxies use to smooth other players

network smoothing mode

there are lot of magic numbers you can tweak to change threshholes for what you consider too far etc

there are quantization levels to sacrifice bandwidth for more accuracy

One thing that helps a lot is to smooth incoming values with interpolation

Unreal does this by default for many things including aim rotation

[limber gyro]

is this in the cmc component BP?

[nova wasp]

yes but it should be visible in Acharacter NetworkSmoothingMode

[limber gyro]

i think i might be looking for this

[nova wasp]

what that means is that if you have two positional updates more than 384 cms apart

it should never smooth between them

[limber gyro]

ye exactly

cms?

isnt it unreal units?

[nova wasp]

generally it can be assumed to be in unreal units

from 2 seconds of reading the source it is indeed just unreal units

[limber gyro]

for reference with my character

[nova wasp]

        const float DistSq = NewToOldVector.SizeSquared();
        if (DistSq > FMath::Square(ClientData->MaxSmoothNetUpdateDist))
        {
            ClientData->MeshTranslationOffset = (DistSq > FMath::Square(ClientData->NoSmoothNetUpdateDist)) 
                ? FVector::ZeroVector 
                : ClientData->MeshTranslationOffset + ClientData->MaxSmoothNetUpdateDist * NewToOldVector.GetSafeNormal();
        }

[limber gyro]

it is quite a lot so there is no interpolation in that distance

it might be what makes it look choppy

this one, if its further away than this its gonna cause a teleport?

[nova wasp]

It's easier to just read what smoothclient does in code tbh

with that one it appears to be a kind of max distance you can be from the next target endpoint during an interp

but I haven't read what it actually does

[nocturne quail]

why an object is auto destroyed on client when scaling in a NetMulticast event?

[nova wasp]

scaling?

like setting the transform scale?

[nocturne quail]

yeah

[nova wasp]

I would need more info to know what's going on here... is the object replicated?

you can probably just breakpoint the object being destroyed and see the callstack for why the net connection decided to send a destroy command

[nocturne quail]

yeah is is replicated and i am scaling it using timeline

[nocturne quail]

i can share the code in a moment

[nocturne quail]

scale is applied in the progress function of timeline

[nova wasp]

looks okay to me... override OnEndPlay and breakpoint it

[silent valley]

Setting visibility to false on the server ends up destroying the actor on clients. IIRC it's the same for a scale of 0.

[nova wasp]

I do not like making a new ShrinkCurve object with nothing keeping track of it... does the timeline reference it for GC?

the scale can scale the timeline and float sample

of an asset ptr curve float instead of this dynamically created one

[nocturne quail]

yeah i made uproperty

[exotic wasp]

setting visibility to false disables replication right?

[lament flax]

I don't see why it would

Visibility isn't replicated iirc

[dark parcel]

it is

you just have to tick component to replicate

actually not sure about Actor vissibility, was thingking of component's

[thin stratus]

@lament flax @dark parcel

AActor has no Visibility. Only USceneComponent.
AActor does have bHidden, which is used in IsNetRelevantFor. So setting that to false (in addition to the RootComponent having Collision disabled!) would cause it to stop replicating.
I can't find anything relating to Scale or Visiblity of the RootComponent affecting replication.

@nocturne quail TimelineProgressUpdate.BindUFunction(this, FName("TimelineProgress")); if you need to use BindUFunction instead of e.g. BindUObject, you might want to use GET_FUNCTION_NAME_CHECKED(ADangerZone, TimelineProgress) instead to at least get the sanity check of not typoing it.

On top of that, it might be better to simply replicate the state of the DangerZone instead of using a Multicast to start it locally. That would ensure that it stays in sync. Would need some form of smoothing though, cause the replicated state could come in choppy as usual.

DangerZoneSphere->SetWorldScale3D(FVector(InitDiameter, InitDiameter, 10000.0f));

Is InitDiameter a Scale value? feels like you are making this actor quite too big.

Also kinda confusign what is going on with the initial Scaling and the FinalScale.

[nocturne quail]

this covers the whole map, and the diameter is 3000.0f sets in the bp actor
its EditDefaultsOnly var

[thin stratus]

DangerZoneSphere->SetWorldScale3D(FVector(InitDiameter, InitDiameter, 10000.0f));

FinalScale = FVector(InitDiameter, InitDiameter, 0.0f);

StartScale = DangerZoneSphere->GetComponentScale();

You are scaling the Z component from 10000 to 0. Is that on purpose?

[thin stratus]

Right, but 3000 is a scale in that function.

[nocturne quail]

yeah its the z value 10000.0f and a constant will never be changed

[thin stratus]

Well your final scale has 0 in it

So it is being changed

DangerZoneSphere->SetWorldScale3D(FVector(InitDiameter, InitDiameter, 10000.0f));

FinalScale = FVector(InitDiameter, InitDiameter, 0.0f);

StartScale = DangerZoneSphere->GetComponentScale();

const FVector NewScale = FMath::Lerp(StartScale, FinalScale, Value);

DangerZoneSphere->SetWorldScale3D(NewScale);

[nocturne quail]

later it is handled

[thin stratus]

Ah in StartShrink

[nocturne quail]

yeah

StartScale = BlueZoneSphere->GetComponentScale();
FinalScale = FVector(NewFinalDiameter, NewFinalDiameter, BlueZoneSphere->GetComponentScale().Z);

on server it works perfectly

[thin stratus]

You wrote the Actor gets destroyed. Is it also destroyed on the Server?

[nocturne quail]

on client it get destroyed on second phase

[nocturne quail]

nah, on server not destroying

[thin stratus]

Did you override EndPlay and breakpoint it o see where the destroy call comes from?

[nocturne quail]

still compiling

[thin stratus]

What is the default of CurrentActivePhase?

[nocturne quail]

this used to read phase data from phase array

[thin stratus]

Doesn't answer my question.

What is its default value?

[nocturne quail]

int32

[thin stratus]

That's the type

[nocturne quail]

default is 0

[thin stratus]

i want to know the value

Ok

That seems wrong then though or not

Cause you ++ it directly on BeginPlay

Means you actually start on Index 1

[nocturne quail]

yeah its fine

tested with default val -1

still destroying

[thin stratus]

And then here

void ADangerZone::TimelineFinished()
{
    if (HasAuthority())
    {
        CurrentActivePhase++;
        if (CurrentActivePhase > ShrinkPhases.Num())
        {

After the TImeline finished teh first time, you would increment to 2, but 2 is not greater than 2, so it will try to read Index 2 from your Array

The if is thus wrong

Needs to be >= or not?

Your Array only has 2 entries, so the indices are only allowed to be 0 and 1.
So yes, you gotta start the value at -1 if you increment it directly and you gotta test against >= Num

[nocturne quail]

i changed it wo strict test later CurrentActivePhase == 7 return

so it never reach outof bounds

[thin stratus]

I can only comment on the code your shared

And there it's wrong

But the EndPlay breakpoint will tell you more

[nocturne quail]

i have actually 7 phases, so yeah according to the code i shared it is now CurrentActivePhase > 1

[thin stratus]

Even with 100 phases, if you Increment the Phase Integer on BeginPlay, it needs to default to -1.
And the test needs to be >= Num

Another thing that is not really great is your usage of Multicast + Replicated Variables

You set all the Replicated Properties and then Multicast StartShrink.

There is no guarantee that the variables arrived before the Multicast.

If you need that, you gotta send them directly via the Multicast instead of replicating them

Preferably neatly packed into a struct.

/// Bad
StartPosition = GetActorLocation();
FinalPosition = NewFinalPosition;
StartScale = DangerZoneSphere->GetComponentScale();
FinalScale = FVector(NewFinalDiameter, NewFinalDiameter, DangerZoneSphere->GetComponentScale().Z);
MulticastStartShrink(NewActivePhase);

/// Better
FDangerZoneShrinkSettings Settings;
Settings.StartPosition = GetActorLocation();
Settings.FinalPosition = NewFinalPosition;
Settings.StartScale = DangerZoneSphere->GetComponentScale();
Settings.FinalScale = FVector(NewFinalDiameter, NewFinalDiameter, DangerZoneSphere->GetComponentScale().Z);
MulticastStartShrink(NewActivePhase, Settings);

There is also the point of the second Multicast potentially arriving before the Client even finished the TImeline.
If the first Multicast got send with a higher ping and the second one with a lower, the second Multicast will arrive before the Tmeline finished.

But I don't think that could actively cause any trouble here, as you fully restart the Timeline and only the server calls the Finish delegate.

Still worth noting as people tend to forget about this shit

You should 100% change your setup with the replicated variables though. That's not just a tip but a bug hazard

[nocturne quail]

I recently changed it to that

    if (HasAuthority())
    {
        ShrinkTimeline->PlayFromStart();
    }

so now everything is in server controll, and also the endlplay is overrided
lets see what is calling the destroy
compilation is finished

testing now

call stack loading

[thin stratus]

Yeah that works, but you probably still want all the different values to arrive at the same time, so if your Timeline drives replicated variables, it's best to bundle them into a struct.

And then use an OnRep of that struct to drive the visuals of the DangerZone

You also gotta realize that this whole thing is a STATE and the Multicast is actually just totally wrong

A late joiner or anyone who needs to listen to the Multicast again, won't get that and will never start shrinking the zone

This should be set up to be fully Server driven, with the Client only receiving the minimum of information to drive the visuals via an OnRep.

[thin stratus]

Yeah so it gets destroyed due to something related to replication.

[nova wasp]

surely not relevancy, right?

[thin stratus]

DistanceBased Relevancy could be it given the randomized location of the actor @nocturne quail @nova wasp

Make sure to mark the Actor as AlwaysRelevant

[nocturne quail]

Thanks it works, no more destroying
so yeah AlwaysRelevant was a thing needs to be true 😄

[dark parcel]

show more

[nocturne quail]

its just a sphere shrinking 😄 using timeline
the code is shared above already

U can generate it in like 10 minutes

[dark parcel]

im interested in your arma mod

[nocturne quail]

its not a mod, its made from scratch 😄

[nocturne quail]

one more capture

seems very smooth now

[dark parcel]

Are you still using multicast?

If you are wat about late joiners

[nocturne quail]

yes multicast, and late joiners are not allowed 😄

[dark parcel]

Well don't forget players outside relevancy

You probably want to make them always relevant too

But that will come with draw back for large map I guess.

[nocturne quail]

I collect 300 of them in the lobby, and put them in Titanic Ship and land them in a different zone 😄

[dark parcel]

battle royal?

[nocturne quail]

then all run away and find loot

[nocturne quail]

kind of yeah

[lament flax]

is there a "simple" way to have more CompressedFlags in the CMC saved move ?

i found https://docs.google.com/document/d/1UO6Ww6Lfpti3YElVdo9uioTUtQJQ9CoSLvd9kF8hvJo/edit?usp=sharing but its heavy

[opal pulsar]

cant you just inherit the saved move struct and put your own stuff ontop

[lament flax]

it still looks like it uses compressed flags

[opal pulsar]

so add another int for more flags and correctly ensure they are serialized

[odd iron]

Hi,
I have a question guys
Is there anyway that I can spawn players on multiple server instances but they can keep presistant and see each other ?
And interact with each other ?

I want to achieve a virtual event like Fortnite one but with big amount of players can see each other

[little pumice]

Isn't multiple servers approach have issues when a lot of players gathers in same location? I mean the key requirements for server meshing like Ashes of creation uses is to have players distributed across multiple zones/grid cells so it can be handled by different "server instances". (not an expert)

[odd iron]

I was thinking about the server meshing but i missing the concept, like how can i make them see each other

[little pumice]

IDK what Fortnite's events is capable of, but I assume smth like music concert or similar art performance are fully scripted from start to end, so it can be played independently on different servers i.e. players on server A and on server B might experience same show without communication between A and B. Though if performance is live i.e. some special player do actions/inputs then it should be replicated to all other server, also if viewers have some way to impact show it will make things even more complicated cuz these actions should be networked to the main server

[lament flax]

I still dont know how Fortnite broadcasted LIVE a show from time squares

Its was sadly a bit low res

[odd iron]

@lament flax @little pumice I have an idea for this but I tried to ask here maybe there is another way to do it.
but the idea i have .
adding the event to the level and connect it with NodeJS or any backend server this server controlling and syncing all instances for example you will start play in the backend server to start a music it will start in the server and it will send the current duration and current progress to all running instances this way it can be synced.
I hope there is other idea xD

[lament flax]

i got no clue sorry

[thin stratus]

I think the issue is mostly that UE servers aren't meant to work like that.
In your scenario you'd need a third party that actually communicates between the servers and sends the locations and what not

What you are suggesting would work fwiw

Some socket connection between a MasterServer and multiple UE servers. Does of course add even more latency

And you'd need to handle all the spawning and destroying of the other players that aren't actually connected to the given server

Sounds like replacing UEs replication system, at least if you want to have a complex version of this, is the smarter idea

Not sure what will come from fortnites side

[odd iron]

@thin stratus Thanks for suggestion ..
I was looking for way or method won't cause high latency and keep maybe 10 Instances in sync or at least they can see same event at same time.

[quasi tide]

At least...not until UE6

[hybrid zodiac]

Hi all, hopefully a straight-forward question: Is it possible for a ReplicatedOn function to be fired before BeginPlay is called on an Actor? I think the initial bunch can be sent before, so all replicated variables might already have been synced before BeginPlay is called, but will OnRep functions trigger before then as well?

[thin stratus]

Yes

Will often not break anything, but if you happen to use a function in that OnRep that requires the Actor the have BegunPlay, then it would cause a bug.

At which point you'd want to call the method by hand on BeginPlay and add an if (HasActorBegunPlay()) check at the start of the OnRep (or inverted with early return)

[hybrid zodiac]

@thin stratus Thanks! I have gone with a boolean called bInitialStateSet which is set to true when the onrep function is called. Then in BeginPlay I only need to "manually" call OnRep if the variable has been changed from the default but bInitialStateSet is still false (i.e. the variable was replicated but OnRep wasn't fired because it was too soon)

[quasi tide]

Why did you go that route rather than the route Cedric said?

It is practically a built-in version of what you're doing.

[hybrid zodiac]

@quasi tide the reason is I want the onrep to run as normal every time, I just need to handle the edge case where the value was replicated without onrep being invoked on the initial bunch, which seemed to be happening

[thin stratus]

But HasActorBegunPlay is handling the edge case

[quasi tide]

I know the situation, but my question is - why did you decide to make a seemingly redundant boolean

[thin stratus]

I mean I guess the BeginPlay call would be unfiltered in that version.

But in a lot of cases you can determine if you want to call it based on the actualy replicated variable

And it also doesn't really hurt in most cases to call it anyway.

void AMyActor::BeginPlay()
{
  Super::BeginPlay();

  if (IsValid(ReplicatedActorVariable))
  {
    OnRep_ReplicatedActorVariable();
  }
}

void AMyActor::OnRep_ReplicatedActorVariable()
{
  if (!HasActorBegunPlay())
  {
    return
  }
}

Something like that, if you really want to filter.

[hybrid zodiac]

Maybe I misread, but the solution proposed was if I don't want onrep to do anything if begin play hasn't been called. I actually want it to always fire and do it's thing, but I noticed occasionally it will be missed. So this is about detecting when onrep failed to fire on initial bunch and manually calling it

In beginplay

So kind of the opposite issue

[thin stratus]

But that's kinda what it's doing

[hybrid zodiac]

Oh I see

[thin stratus]

Sure it will always go that path with this example, but it still solves the issue

[hybrid zodiac]

Yes sorry

Your code snippet made it clearer what it's doing

Does this not mean though that it might be double-called?

[thin stratus]

What you wanted to do is only call in on BeginPlay if it didn't call at all.
While the above solve that and it calling too early

No, cause of the HasActorBegunPlay condition

[hybrid zodiac]

Yeah OK I see

Sorry, I misunderstood your original suggestion

[thin stratus]

No worries. Your solution is slightly different and has a slightly different effect fwiw

[hybrid zodiac]

I'll try your way in my next branch

For full context, when a player joins a game their state is set to "just joined" and it displays a welcome widget on the screen. The state is initialised to "none" in the constructor. Hence why I need to ensure it's always onrep'd otherwise no welcome screen appears, but I don't want it double-called so it wastefully clears and redisplays the widget twice

But your way makes sense too without the manipulation of booleans

Thanks @thin stratus and @quasi tide , I just misunderstood the original suggestion 🙂

[errant charm]

in a listen server, how can i open a level without level streaming and load all the players into it?

or would i have to use sub levels

[modest crater]

I recall a while ago there was whispers of new RPC types coming - or perhaps it was a PR I dont remember exactly but does anyone know what happened with that?

Such as multicasts for proxies for example

Here it was https://github.com/EpicGames/UnrealEngine/pull/11578, still open

[quiet yarrow]

maybe data layers?

[errant charm]

honestly just going to use level streaming

[quiet yarrow]

So I cant figure out for the life of me why my attach actor to component wont work with network emulation on

It works perfectly with low ping

but in high ping the object attaches, but its location is wildly off for no reason that I can find

[dark parcel]

Don't know what sort of attachment but why don't you do the attachment locally?

Only thing you replicate the is the object you are attaching

[quiet yarrow]

thats true

[dark parcel]

It might be that the attachment takes the info from the server.

When arrived at client since there is delay

The position wouldn't be the same

Kinda like running a line trace on every machine

You won't get the same result

[quiet yarrow]

but

[dark parcel]

If you attach to a socket or bone, there shouldn't be issue though?

[quiet yarrow]

unless I update the location manually if the other clients never knew about the attachment how would they know the location of the object?

[dark parcel]

What sorts of attachment is it

A weapon?

[quiet yarrow]

actor component

[dark parcel]

Accessories?

[quiet yarrow]

its just a battery

purely visual

[dark parcel]

Actor component don't have physical presence, they don't have transform

Static mesh comp perhaps?

[quiet yarrow]

yes sorry, thats what i meant

I could try attaching to a socket

ill let you know if that works

it is a staticmeshactor

[dark parcel]

Ye attaching to world position would give you different result most likely.

By the time the location is replicated the actor would be in different place.

The higher the ping the more apparent the distance is

What you can do is probably pick a point from the parent mesh and set the location relative to the point (sockets / bone) perhaps.

Or just relative to a component location

[warm roost]

Has anyone tried to make multiplayer in VR? What online subsystem could you suggest for VR multiplayer over internet (not LAN)?

[quiet yarrow]

looks like setting the actors relative transform after the attach actor fixed the issue. I guess attach actor snap to target isnt good enough

[sonic obsidian]

hey all looking for some advice on net relevancy for my foliage actors like trees etc.. im wondering if i should make them always relevant so players in the distance can still see the actor being destroyed these foliage actors arnt replicating movement or anything just abit concerned about the network performance with this as i have a large amount of those actors

[hoary spear]

Consider value to the player vs the cost.

Is it likely the player will even notice this?

[kindred widget]

I was about to make a point that you'd probably hit GPU issues before network ones, but then I remembered that everything is Nanite anymore.

[chrome bay]

If I wanted to destroy foliage I definitely wouldn't replicate those actors directly either way.

wayyyyy too much for networking to deal with

[weak plover]

I have a portal system taken from Lyra that works fine (replicated) when on the same level. But when a player (client) goes through the portal and it leads to another level, all other players (both client and server) is also teleported to the new level lol.** Anyone have any good tip on this or where i can learn about the topic?**

[nocturne quail]

is there any function like FindCrowdedSpotInMap() 😄 ?

if not how i can implement it?

my guess is to store the locations of all players in an array, and then look if the more locations with less difference in distance should be that spot

FVector FindMostCrowdedSpot(const TArray<AArmaCharacterBase*>& Players)
{
    FVector MostCrowdedSpot = FVector::ZeroVector;
    int32 MaxPlayerCount = 0;

    for (const auto& Player : Players){
        int32 PlayerCount = 0;

        for (const auto& OtherPlayer : Players){
            if (Player != OtherPlayer){
                PlayerCount++;
            }
        }

        if (PlayerCount > MaxPlayerCount){
            MaxPlayerCount = PlayerCount;
            MostCrowdedSpot = Player->GetActorLocation();
        }
    }
    return MostCrowdedSpot;
}

void ADangerZone::TryNukeCrowdedArea()
{
    if (!NuclearBombClass) { return; }
    
    NotifyAllClients("Attention: Crowded area will be nuked in next 3 minutes...");
    
    TArray<AArmaCharacterBase*> Players;
    FActorSpawnParameters SpawnParams;
    SpawnParams.SpawnCollisionHandlingOverride = ESpawnActorCollisionHandlingMethod::AlwaysSpawn;
    FTransform BomberTargetTransform;
    FVector MostCrowdedSpot = FindMostCrowdedSpot(Players);

    BomberTargetTransform.SetLocation(MostCrowdedSpot);

    if (ADropNuclear* BomberTarget = GetWorld()->SpawnActor<ADropNuclear>(NuclearBombClass, BomberTargetTransform, SpawnParams)){

    }
}

[blazing bear]

To those who have worked with AI, do you run logic in the pawn class of the AI as well? And if so, only from the AI Controller calls to make sure it's ran by the server right? I'm trying to understand what's the relationship between the pawn and ai controller, but as far as I've gotten to understand it, is that the AI Controller is the authority, besides setting the blackboard values for the behavior tree, it also tells the pawn and calls its functions right?

I'm saying this because I've tried putting logic in the pawn and have it ran by itself but apparently the pawn in the client's machine is owned by the client so there is no server/client relationship when it comes to solely to the pawn

[chrome bay]

Controller executes the AI

pawn is the body, controller is the brain

AI always runs on the Server only

[blazing bear]

gotcha, makes perfect sense.

thanks

[crisp shard]

curious on ppls opinions, is a dedicated server and save games enough for persistent storage of a player's data ? and if not, what are some simple options i can look into ? (hopefully dev friendly)

[chrome bay]

you can't store save games on a dedicated server no

you'll need some kind of persistent backend of your own, or use an existing service

[crisp shard]

i figured, i had an SQL database workign for a while and that was going to be my answer to the important data, but it doesn't scale well (at the least the way i was doing it). it wasn't really possible to load all the data at once, would cause crashes and other issues. might have also just been the particular plugin i was using to faciliate it

what are some semi user friendly exisiting services ?

i've heard of playfab, never looked into it, i think that's for this

[chrome bay]

not aware of any but I'm sure some exist

EOS might have it too. Try #epic-online-services

[crisp shard]

yea im actually using EOS and i do know they have something like it but i have heard it's not the most reliable, as in i think it still is subject to tampering

[twin juniper]

How can I create a mod for unreal engine 4 that adds a multiplayer game if the game is single-player?

[keen adder]

You can’t. To get multiplayer working you’d pretty much need to redesign the entire code architecture. Ex: you cant use returns on online functions.

Unless you have access to the source code and want to essentially remake the game

Even bungie said they couldnt make halo 1 online multiplayer when xbox live released because they would have been essentially remaking the whole game — and they were the creators lol

[lost inlet]

Also #rules since I assume this is unauthorised modding

[sharp hamlet]

hello, this DEF variable is giving me "0" value which is the default value. but at character begin play i m setting the DEF value appropriately which I confirmed by using print string. pls help

[thin stratus]

Print the Value on BeginPlay and in PointDamage and print the DisplayName of the Blueprint (self) in addition to check if you are really setting it correctly.

Also make sure this is called only on the Server (Point Damage).

[sharp hamlet]

i checked i used a print string after this function, its returning me "0" but if I printstring from a debugkey, its showing me appropriate value

[sharp hamlet]

PointDMG is server only

getting "0" in this print string

getting appropriate value in this print string

[lament flax]

oh damn, mantling sounds awful to manage for decent networking

same for rootmotion

[exotic wasp]

that's a rule?

[lament flax]

nb 5

[exotic wasp]

not even going to question how modding==piracy

[thin stratus]

Please print the name of the BP (self) alongside that.

There is a good chance you are printing the values of 2 different instances for whatever reason.

[thin stratus]

You mean mantling on top of obstacles?

[lament flax]

yeah stuff like that

rootmotion stuff

[graceful flame]

How / when are you setting the value of DEF? Is the value being set on the server then replicated over the network to the client? If you set it on the client it won't be replicated upstream to the server.

[sharp hamlet]

Yes i m setting on the owning client

[thin stratus]

Why would that be problematic?

[graceful flame]

Right so that explains it

[thin stratus]

Especially if you use the CMC it should be fairly straight forward.

[graceful flame]

Replicated data only travels down from server to client never the other way around.

[sharp hamlet]

So I shud server->mc that?

[thin stratus]

No, the value is marked as replicated. You need to set it on the Server, that's it.

[graceful flame]

Don't need to multicast

[thin stratus]

Given the Actor itself is marked as replicated.

[sharp hamlet]

Let me try this. Thanks a lot

[thin stratus]

If you still have issue, please share the code that sets the DEF variable value.

Currently you only shared the part where you use it.

[sharp hamlet]

Yessssss

It workssas

Aaaaaaaa

@thin stratus & @graceful flame thank u so so much

Sry fr image quality, my ethernet is gone

This is where I was setting the value

[thin stratus]

Right okay. I would suggest you to learn a bit about inheritance, cause what you are doing there doesn't scale at all.
These different stats should be set in a Child Class (e.g. BP_Character_Lucy).

And then just in the default values and not in runtime code.

[sharp hamlet]

Yea just like weapons right?

[thin stratus]

Yeah, anything that sort of gets more specific down the tree.

Weapon -> Melee -> Sword1
Weapon -> Melee -> Sword2
Weapon -> Ranged -> Bow1
Weapon -> Ranged -> Crossbow1

[sharp hamlet]

Would u suggest that I fetch it from a savegameobj?

[thin stratus]

Character -> Lucy
Character -> Riot
Character -> XYZ

[thin stratus]

If those stats can change in the future then you'd at some point save and load them yes

[sharp hamlet]

I have a master bp for weapons, i have to create one for characters as well

[thin stratus]

The can totally be done during runtime, but for the actual base values, you'd just set them as default values in the different characters.

[thin stratus]

You usually do that for a lot of things.

Object Oriented Programming makes a lot of use of Inheritance

Doesn't always need to be one single master BP either

Like with the Weapons, some more specific code can live in a Melee and a Ranged child

From which then the more specific child class inherit

Most people have something like this:

• BP_Character
  • BP_PlayerCharacter
    • BP_PlayerCharacter_Lucy
  • BP_EnemyCharacter
    • BP_EnemyCharacter_Skeleton

[sharp hamlet]

I actually am fetching the character mesh frm savegameobj. I have to rewrite that whole code to fetch from the child character bp?

[thin stratus]

To ensure that general Character stuff can live in BP_Character. And Player vs Enemy is split.

[thin stratus]

Can't really answer that. I don't see why the Character Mesh would live in a SaveGame tbh

Unless BP_PlayerCharacter_Lucy can have different Meshes for the same Lucy character.

[sharp hamlet]

Noob programming on my part in the very start, but theres a lot of code after that in the mainmenu apart from savegameobj

I will rewrite this. Thank u

[thin stratus]

Usually you code most of your game before adding the SaveGame stuff.

The SaveGame stuff should just be a minimal version of information to recreate the state of the game.

If your Character has any runtime data that changes, like health or so, that would go into it.
If the Mesh of the Character is set it stone due to it being BP_Character_Lucy, then that doens't need to go into the SaveGame.
What would go into the SaveGame is that the player is using BP_Character_Lucy in the form of the Class of the Blueprint.

[sharp hamlet]

So I should not save info in the savegameobj from a http get request from a server? But I also wrote a code to validate the values too

[thin stratus]

I mean, SaveGames are usually for.. saving the game.

[sharp hamlet]

This cleared my confusion

[thin stratus]

If you save the game into some backend server with some HTTP request, then that doesn't really change WHAT you are saving.

You'd still save the Health, and the BP Class

And the Stats

And the Armor that is equipped, etc.

Doesn't really matter if SaveGame or HTTP Request to a Database.

Combining HTTP Request and SaveGame can make sense if you want to support offline gaming and sync it with the backend later.

[sharp hamlet]

Not my aim

[thin stratus]

But that's then more about your save game stuff and not really about the original point of Inheritance anymore

[sharp hamlet]

So I shud make a new system for these stats by using a master bp, and different characters in its child. Am I right?

[thin stratus]

Yeah, just don't think about it as ONE Master BP.

You can branch off a bunch of times if you want to have specific grouping of code

E.g. Player vs Enemy

Melee vs Ranged (Weapon)

[sharp hamlet]

Pls i didnt understand this

[thin stratus]

You don't need to do this:

• BP_Character
  • BP_Character_Lucy
  • BP_Character_Riot
  • BP_Character_XYZ
  • BP_Character_Skeleton

You can (and often should) do this:

• BP_Character
  • BP_PlayerCharacter
    • BP_PlayerCharacter_Lucy
    • BP_PlayerCharacter_Riot
    • BP_PlayerCharacter_XYZ
  • BP_EnemyCharacter
    • BP_EnemyCharacter_Skeleton

You don't have to instantly split it like this, but you should keep it in mind.
There might be code that is only relevant for Player Characters, and there might be code that is only relevant for Enemy Characters (AI code for example).

And if it makes any sense, you can add even more in-between classes.

[dark parcel]

Probably want to understand OOP before knees deep into your project

[sharp hamlet]

Thank u so much for explaining.

Rn, my code is this:

in mainmenu, in char selection, select char and save to savegameobj.

Ingame, fetch savegameobj and get the mesh, then set mesh

[thin stratus]

Yeah, I see, you are trying to persist the data between MainMenu and Gameplay map.

And that's also totally fine. Just the stuff in the SaveGame could just be the Character BP Class

And BP_PlayerCharacter_Lucy has the default values for the stats then.

[sharp hamlet]

And this helps the user to load the game and have their previously selected char to be currently selected

[sharp hamlet]

Yes, this is what i shud do

[thin stratus]

Yeah it can be used for that to

[sharp hamlet]

I have this system for the weapons though. Will make one for the characters

[limber gyro]

does the server have interpolation at all?

[grand kestrel]

@thin stratus I'm unable to get the same result as compressed flags with move containers. They result in unrecoverable de-sync in rare cases. Do you have something like a basic sprint implementation that uses move containers instead of compressed flags? I feel like I must be missing something obvious

[dark parcel]

What's the context?

[limber gyro]

just moving side to side

[thin stratus]

The majority of my last year was Mover. I don't have anything at hand for CMC, sorry.

Unrecoverable de-sync sounds bad though, that shouldn't happen.

The Container should be the same as the Flags.

Old setup send the Flags via the ServerMove call. New setup send the container struct.
That's kinda all there is.

[grand kestrel]

I'll keep looking

[thin stratus]

The Flags might be in the SavedMove of course, to replay moves, but I assume you added whatever booleans you set in the container for the input flags also to the SavedMove.

[grand kestrel]

In theory if I replace the compressed flags with move containers nothing else should be required?
https://github.com/Vaei/PredictedMovement/blob/main/Source/PredictedMovement/Private/Sprint/SprintMovement.cpp

[dark parcel]

Are you having particular issue? Movement pretty much handled already and you shouldn't have jitter or snappy movement if everything is set right.

[limber gyro]

it just looks jittery on the listen server

and from what google tells me im not the only one noticing this

[grand kestrel]

It would be exceptionally difficult to repro with sprint tho
I'm testing with a slow fall/levitate instead, and the de-sync happens in really rare cases, likely when one thinks its on ground, other doesn't

[dark parcel]

Something to do with the tick pose

[limber gyro]

https://issues.unrealengine.com/issue/UE-32005

[dark parcel]

Kaos settings help migitate the effect.

[limber gyro]

why the heck wouldnt the listen server have interpolation?

[thin stratus]

In theory you would need to:

• Put the Flags into the Container
• Add something that works like USprintMovement::UpdateFromCompressedFlags but for the Container
• Add something that works like FSavedMove_Character_Sprint::GetCompressedFlags but for the new flags from the Container

The flags aren't that magical that they could break something and cause a de-sync, imo.

FSavedMove_Character_Sprint saves a boolean for FSavedMove_Character_Sprint::bWantsToSprint and sets the Flag in GetCompressedFlags if that boolean is true.
USprintMovement uses the Flags to set USprintMovement::bWantsToSprint.

CMC uses FSavedMove_Character::SetInitialPosition to set bJumpPressed instead of SetMoveFor. But it does use SetMoveFor for the Crouch stuff.

Not sure if your specific stuff there should do anything with ImportantMove or bForceNoCombine.

[thin stratus]

Main reason is that the Server doesn't tick the Pose of the SkeletalMesh on Tick, but rather with each received ServerMove RPC.
Reason for doing that, from Epic's side, is to keep the Animations in sync with Movement iirc.

Doesn't really help with the fact that it does indeed make Clients look laggy on the Server

[limber gyro]

so is there anything u can do?

i mean, its not the end of the world in custom games

but still

[dark parcel]

Kaos post .ini setting and it helps

Going dedicated solve this specific issue though

[thin stratus]

A few years ago the usual option was to inherit from CMC and Character in C++, override the parts in Character that set the ticking to manual back to automatic and remove the parts in the CMC that tick the pose manually.

But the CMC seems to have gotten more complex in that (who would have thought) and I'm not sure it's that straight forward anymore.

I'm pretty sure The Ascent shipped with that fix :D

[lament flax]

i dunno i feel i will get some desync

how much does CMC handle anims/root motion anims ?

[thin stratus]

Puh, not so straight forward to explain.

It doesn't handle Animations. Only RootMotion and RootMotionSources.

And the RootMotion it usually grabs from AnimMontages that are currently playing.

RootMotionSources is something you mostly get into contact with when using GAS, but you can also use them outside of GAS (just Epic never exposed them).

If you play a Montage with RootMotion on Server and Client, then it will keep you in sync.

[lament flax]

okay, so i need to play them on both

multicast incoming

[thin stratus]

I mean, we are talking CMC and predicting usually.

So probably local + ServerRPC ? :D

[lament flax]

yeah ill do that

thanks for the headsup

[vapid gazelle]

I got a dummy question for y'all. Say I'm implementing a game like Rocket League where I have a ball that's being kicked around by a bunch of players, and I want the ball's movement to feel buttery-smooth on every one of my clients.

My impression is that if I just naively say "the server will tell you the position of the ball every step of the way" then inevitably the ball will be jittery as heck for everybody, my client could be running at 240fps but I may be only getting 30 transform updates per second for the ball, no good.

What's a practical solution for this in UE? How do developers go about making sure the bouncy ball being whacked around looks great for every player? My guess is that every client locally has to apply the same impluse on the ball and locally simulate its physics and trust the server to keep it honest at regular intervals, but... what does that actually look like in practice?

[fossil spoke]

Replicated Physics movement of anything is extremely difficult.

Im not entirely familiar with Rocket League or how they manage it.

But I wouldnt be surprised if they did big modifications to the engine to setup some sort of deterministic physics across the network.

The easiest solution which will get you close but still look somewhat rough, would be to smooth between the network updates.

[vapid gazelle]

Sounds like there's nothing obvious off-the-shelf in UE that is magically plug and play and will give me this sort of functionality I'm looking for, right?

[fossil spoke]

No

[vapid gazelle]

There's this talk from several years back which I need to revisit, but I was mostly digging around to see if perhaps since then something new showed up in the toolbox that solves it. IIRC Rocket League is on UE3.

https://www.youtube.com/watch?v=ueEmiDM94IE

[fossil spoke]

Supposedly Chaos Physics is meant to be able to manage this type of thing in the future.

But its a long way off from my understanding.

[swift sorrel]

from my (limited) research, there is no easy way to get replicated physics. you will always be fighting against server-authoritative network prediction with CMC/any other built-in movement system you're using. you'd have to create your own framework pretty much

[dark parcel]

Pubg doesn't even predict it.

The ball got kicked 1 second later

Iirc the dudes from rocket league made video and speech on how they do it

You may get the overview but it's not gonna be a simple steps

[limber gyro]

could he get something similar if he treated the ball as a projectile?

projectiles are syncd between everyone aren they?

[little pumice]

BallCharacter

[subtle kernel]

hey guys, I notice that once my actor becomes irrelevant(IsNetRelevantFor returns false) it will be destroyed from a client with a delay. And delay seems to be bigger then my network latency. Is it so? Is there any way to control this delay? Thanks

[fossil spoke]

The delay is on purpose. It is a buffer, this helps reduce the potential for the Actor to constantly be switching between being relevant and not, which would be expensive.

It cannot be changed.

At least not without modifying the engine.

[subtle kernel]

I'm just using relevancy as one of my game mechanics for prototype, not sure if it's a good idea though 🙂 but I will give it a try

[fossil spoke]

I would not rely on relevancy as a game mechanic.

You should setup your own soft form of mechanic that mimics its effects.

[subtle kernel]

yea, I will do it later, don't want to waste time on it right now

[blazing maple]

One message removed from a suspended account.

[nocturne quail]

This !HasAuthority() check will also avoid something to be executed on dedicated server or i have to check for if (GetNetMode() != NM_DedicatedServer) ?

my guess is that !HadAuthority is enough

but not sure

[nova wasp]

listen server players have authority over replicated actors

There's also FApp::CanEverRender I guess

but generally you just GetNetMode to determine if it is dedicated

[nocturne quail]

Solved lots of confusions thanks

[nova wasp]

if you will NEVER have listen servers you can ignore this, but I recommend you keep listen servers working just for the sake of how useful it is to see authority things rendered

[nocturne quail]

i will do

if (GetNetMode() != NM_DedicatedServer || !HasAuthority())

[lament flax]

or compare GetNetMode() with a int

if GetNetMode() < NM_Client

[chrome bay]

This. It's actually not as hard as people think to support both too

[nova wasp]

It's not but it does add a bit of an extra situation to consider in some larger manager things

but even if you can just keep like... the cmcs moving around etc it will still be useful to see what is happening

[chrome bay]

Yeah, if you can support both you defo should. There's no substitute sometimes for being able to actually see what's happening server side

[median igloo]

im trying to make local multiplayer for a school project, when spawning my second player it doesnt have any controls :) any pointers?

also for some reason another copy of player controller 0 spawns (the one listed as my default pawn)

[nova wasp]

https://wizardcell.com/unreal/multiplayer-tips-and-tricks/#2-beware-of-getplayerxxx0-static-functions
(woops, they meant local!)

[median igloo]

yeah so what i have going right now is that the Game mode spawns player 0 which works with controlls just fine but player 1 doesnt work at all with controlls

perhaps I have set up the controlls wrong? i copied and pasted the first player bp and changed the keybinds 😂

[nova wasp]

how would player 2 get the second player index controller when they only have one controller on their client

[median igloo]

its local multiplayer, theyre the same client

[nova wasp]

ah, I'm a dingus and missed the "local" part

[median igloo]

im only a first year im not doing all that 😂

[dark parcel]

not a lot of people doing local mp fyi

so it may take some times to get answer

[median igloo]

i know unreal tries to do a lot for you when it comes to lmp but like, it would help a lot if i could just copy and paste the player 1 bp and change keybinds and call it a day

[dark parcel]

when you say it doesn't have any controls, you mean it's not taking input?

have you bind the inputs to the 2nd player?

[median igloo]

yeah

i gave it possession and added the input context yeah

[dark parcel]

how and when do you bind the input mapping context?

[median igloo]

its not in that screenshot i just realise

give me 2 seconds ill open it back up

something i will add

is that the get controller call fails

i think this is because im in my level blueprint, and the level blueprint isnt direived from pawn

however im very much shooting in the dark there

[dark parcel]

can you show the whole bp? I can[t tell when this is executed

btw don['t use level blueprint

also you might be mapping it when the controller is not even spawned yet

dunnoe if that matter, since I never do local multiplayer my self

you probably want OnPosses -> Add Mapping Context

[median igloo]

i just moved the mapping to the player blueprint

thoes are the two blueprints involved

[median igloo]

i dont do C++ yet :( is this a blueprint node too?

[dark parcel]

yes, pertty much what you are doing for player 0

you should do for player 1

imo, I will just move the binding in the OnPosses event

[median igloo]

its spawned by the game mode as its the default char

[dark parcel]

i'm just saying that you might have to bind the input for the player 1 too

imo, you can probably just bind on posses

so regardless which controller get spawned or you want to set.

everytime a controller posses a pawn, they will bind the input.

[median igloo]

and i can just run that in their respective blueprints?

is OnPossess what the node is called? i cant find it 😭

[dark parcel]

it's an event that gets called when a controller posses a pawn

[median igloo]

OH i see

event possessed

found it

[dark parcel]

you probably want to make a function in your character to bind the input if you are using Input Action in the character.

then from the controller, when on possess is called, get controlled pawn -> cast to your character -> bind input

[median igloo]

okay i understand, i will try now quickly

if not i will ask my teacher what the fuck to do

i was at this all night

yeah im still getting nothing from that player, something weird is that even when i make player 2 the default pawn and spawn it its not moving

so i wonder if its an imc problem?

not that theres anything here to go wrong

[dark parcel]

Show your codes

Is player 2 different bp?

[rapid mango]

Hi guys.
I faced issue with OnlineSessions.
I added Custom Settings with code like this :

IOnlineSubsystem* OnlineSubsystem = Online::GetSubsystem(GEngine->GetWorldFromContextObject(WorldContextObject, EGetWorldErrorMode::ReturnNull));
    IOnlineSessionPtr Sessions = OnlineSubsystem->GetSessionInterface();
    
    if (!Sessions.IsValid())
        return false;

    FOnlineSessionSettings* SessionSettings = Sessions->GetSessionSettings(NAME_GameSession);

    if (SessionSettings)
    {
        SessionSettings->Set(InSettingKey, InSettingValue);
        return Sessions->UpdateSession(NAME_GameSession, *SessionSettings, true);
    }

    return false;

After find sessions node in blueprints i call this :

if (!InSession.OnlineResult.IsValid())
        return false;

    return InSession.OnlineResult.Session.SessionSettings.Get(InSettingKey, OutSettingValue);

but SessionSettings.Settings are always empty.
I check with breakpoints and SearchObject in FindSessionsCallbackProxy just don't download them.

I use OnlineSubsystemNull and Lan sessions

.ini :

[OnlineSubsystem]
DefaultPlatformService=LAN

Sessions are downloaded successfully, but without my custom settings.
Any ideas?

[tardy fossil]

try doing [OnlineSubsystem] DefaultPlatformService=Null cause I dont think "lan sessions" is a thing.. but i could be wrong

[rapid mango]

I tried, but this did't help

[rapid mango]

I've got this.

SessionSettings->Set(InSettingKey, InSettingValue);

should look like this :

SessionSettings->Set(InSettingKey, InSettingValue, EOnlineDataAdvertisementType::ViaOnlineService);

[glossy wigeon]

Hey guys, currently I'm doing a Capsule Trace in the GameplayAbility, to check for melee hits, and doing it only if HasAuthority is true.
But I want to move it, maybe into a AnimNotify, but then this would not be executed on the server and so I can't check for HasAuthority there.
What/How would or could I do there, what would be a good solution for that?

[upbeat basin]

Are AnimNotifies not run on both dedicated and listen servers by default?

[glossy wigeon]

I'm not sure, I'm quite new to Multiplayer, I would have assumed that the Animation is not played on the server and hence the Notify is also not done on server.
But if they are executed on server and client this would be awesome!

[upbeat basin]

It should be playing for the sake of notifies at least. There is also an option to disable the notify trigger on dedicated servers (which is enabled by default)

What server should be missing is actually updating the bone locations for the mesh according to animation since it's not being rendered, which is the case for dedicated server only, since a listen server would still render the scene

And often times we did enabled the bone location update on server too. On FPS projects, disabled bone update causes the player mesh to stay in T pose for server, which makes it not consistent between server and client to shoot that player for example

So your animation timeline and events (Montage start/end/blend and notifies) should still run on any kind of server by default

[glossy wigeon]

Wow ok thats amazing, so I could probably just check for authority in the notify and then do the capsule trace right?

Also the info regarding the bones is also really good to know, maybe at the moment I don't need it but when time comes its really good to know.

[upbeat basin]

Yes. AnimNotify itself might not have the auth check node/function since it's a UObject based type but you can get the owning mesh component and it's owner actor to check if you're on server or not

[glossy wigeon]

ohhh i see and then since the animation is also played on the server, it will check there.

[upbeat basin]

It was a painful process for us to figure out why a seemingly valid trace on client was not hitting the character on the server and a simple dropdown option on the SKM component was the fix back in the days. So I try to add this bit of information every time, even if it's not relevant to the actual question

[glossy wigeon]

But the only downside regarding this, if a client hits, and he is delayed or something, and on server the characters are already positioned different, the client would see a hit but it wont register

but i guess this is not to bad and no realy way around that right?

[upbeat basin]

Well yes, this is a whole different topic about prediction and rollback, which I'm also not so much experienced with

[dark parcel]

you can design it how ever you want

the easiest one is obviously just to accept that client did a hit

[glossy wigeon]

Yea I wonder how complicated it is and if it's worth getting into that or maybe just if it hits problems.

[glossy wigeon]

You mean do it client sided and send a event to the server right?

[dark parcel]

yeah you can do that

next level from that is some server validation

[glossy wigeon]

It's just if someone is cheating bad right?

[dark parcel]

server chekc if the hit is possible or not

[dark parcel]

then you want to step up your game

but if cheating is not a concer, there's the path of least resistance

The server validation can be as simple as, checking if there's anything blocking the trace, or if it's close enough.

the other one would be to save Character's hit box every frame

doing some rewinding on the server to see if the hit is possible or valid

[glossy wigeon]

so basically just a way of checking if the hit that the client said it did is not to far of right?

[dark parcel]

how you want to do it is up to you

who has the say and so on.

[upbeat basin]

Yup it all depends on the needs, every solution has it's own pros and cons concerning development time, cpu usage, cheat protection etc.

[glossy wigeon]

yea thats true

I'll maybe try something out like the client solution or doing it just server sided, maybe wit ha capsule trace, there ight not even be such a big issue even for some ping difference

[upbeat basin]

It could be enough if it's not a fast paced competitive fps game or a personal/learning project

You can always upgrade/refactor the logic when you want/are required to step up your game as ColdSummer said

[glossy wigeon]

Yea, I mean if there will be some cheating, the ofc some more logic has to be implemented, but idk.
i mean in lyra the melee hits are only done with authority.

But having a complex system like ColdSummer mentioned, might be safe solution but on the other hand it probably takes quite some time to do and then to get right as well.

but thank you guys a lot for the help and info 🙂

[pale hazel]

Can anyone help me figure out why my listen server seemingly is not properly working?

I create a session and then call GetWorld()->ServerTravel("/Game/Maps/MyMap?listen")
This works so far and the ListenServer loads into the map

But when I now try to find the session and connect with a client, it joins the session successfully, but the ClientTravel fails. The connection times out and it never successfully connects to the listenserver

I use GetResolvedConnectString() to get the address to connect to, and it returns an IP that has :0 at the end, so port 0. I guess this is the issue?

I should add im using OnlineSubSystemNull, and trying to do it in LAN

[upbeat basin]

Are you testing on PIE?

[crisp shard]

so i've successfully setup possession of a pawn and have some controls for it and what not, and im trying to think of best way to go back to original pawn

i have a widget that spawns and possesses the second pawn, this changes the widget to essentially view the pawn via a render target. you can't move , just look around, and do whatever else, but im not certain on best method of going back to original pawn

i found that best way to spawn and possess was doing it via game mode but i dont necessarily want to store the additional pawn on the game mode, as i could see this getting messier than it needs to, by storing an array of pawns / actors to keep track of, but im not sure how else i'd get reference to the controlled pawn as it was never stored as a ref anywhere. not sure if the game mode woul djust be the best place for that, but yea any thoughts would be cool

[pale hazel]

No, I'm testing packaged builds.
So I got it to work, but not sure why:
Instead of using GetWorld()->ServerTravel(...) to open the level on the ListenServer, I use UGameplayStatics:OpenLevel(...), and now it works

[upbeat basin]

Oh I guess if it was causing a problem it would not give you any sessions after find session to begin with

[pale hazel]

Yeah

If you have any insight on why GetWorld()->ServerTravel(...) wouldnt work I'd love to hear it, the Tutorial I was following used it and it worked fine for them 😅

[upbeat basin]

Hmm I feel like I remember a conversation about this

[upbeat basin]

#multiplayer message

I don't have any official statement or explanation to quote but it seems like ServerTravel doesn't change the NetMode for the target world(?)

I am also currently using GEngine->Browse to open the listen server and then ServerTravel to take players from menu to game level

[pale hazel]

Hm, that's a good guess for sure, thanks for the help!
I'm just happy that it works for now haha

[crisp shard]

update: i just stored the new pawn on the controller do the work in the game mode again by sending it to the game mode then calling unpossess event in new pawn from game mode, that works

[glad escarp]

Hey all. I need some tips for multiplayer spawning. I am experimenting with a procedural level generator and even if I weren't, I would like to learn the proper way to handle spawning in multiplayer if there is a best practice beside letting the GameMode manage it automatically.

[haughty ingot]

A lot of times you can just pass the seed you use for generation to the connections that need it, have them generate itself and use stably named actors. I think there’s a blog post about around here somewhere

https://vorixo.github.io/devtricks/procgen/

[glad escarp]

Rad. Thanks.

[glad escarp]

Now to read this and translate to BP

[haughty ingot]

Some stuff won’t be possible without C++ (like forcing actors to be stably named).

[glad escarp]

Guess I'll have to do some C++ classes then.

[glad escarp]

Side-note. I'm trying to test some stuff in multiplayer now but my server and client are not together. I have it set to play in New Editor Window (PIE), Number of players is 2 and Net Mode is "Listen Server" but when I hit play, they do not spawn into the same map together.

I do have startup movies enabled and in my Maps and Modes I've set the defaults to the current map I'm testing instead of my title screen. Not sure if either of those things makes a difference to what's going on in my editor.

[nocturne quail]

enable replicated on character

[glad escarp]

The Player Character is replicated of course.

[nocturne quail]

if you are playing as listen server with two number of clients, one will be server and second will be your client
so server and player should both see each other, at least server will see the client in any case

else you are in standalone mode

check it again

[glad escarp]

I just did. The player character is replicated and I am in PIE not standalone. It loads the same map but the players do not appear to be together or see each other. It's very strange as I have tested everything before and this wasn't happening.

[nocturne quail]

you have a different server map in your maps/mods ?

[glad escarp]

I don't have a server map in there. I was under the impression that was for dedicated servers so I left it blank.

[nocturne quail]

print something when you press 1 key on a multicast event and check what you will see

[crisp shard]

Update: I changed that. Now it’s stored in the original character, so that if dies it knows you’re controlling the other pawn and will switch back so you no longer control pawn.

Any controller related events on char I had to opt out of if the controller is controlling a separate pawn , so again ref on chat helped w this just checking if valid. All other events stay as they seem to work fine

[glad escarp]

The print strings only occur on the individual instances of the game. It prints Server: Hello on the server and then just Hello on the Client but doesn't print Client: Hello on the server or Server: Hello on the client.

[grand kestrel]

Ahem, sorry, ignore that @, Cedric
I was actually 99.9% of the way there

[thin stratus]

I wonder if you want the CanCombineWith code oyu commented out

iirc the parent does that via the Flags != other.Flags

At the same time you don't want to have the CombineWith stuff do anything with bWantsToSprint

This is an important move after all

You only combine if the result of the two moves together is the same as if they are done one by one

But if you start sprinting in one but not the other, then you can't combine them

@grand kestrel

[grand kestrel]

@thin stratus I think you're right, they shouldn't be combined

Move Containers vs Compressed Flags

Where we use GetCompressedFlags(), we now use FCharacterNetworkMoveData::ClientFillNetworkMoveData()
This is where the client packs the inputs (e.g. bWantsToSprint) for sending to the server, and there is a matching required Serialize() function as well.

Where we use UpdateFromCompressedFlags(), we now use UCharacterMovementComponent::ServerMove_PerformMovement() (Super should occur after your code!)
This is where the server unpacks the inputs that was sent from the client

So long as you replace these two functions, the remainder of the predicted movement code can be identical to using compressed flags!

You can do more, like having the server send the client data by using FCharacterMoveResponseDataContainer::ServerFillResponseData() and implementing the matching Serialize(), and the client receives this in UCharacterMovementComponent::ClientHandleMoveResponse()

You can further use UCharacterMovementComponent::ServerCheckClientError() to compare the difference between server and client and force a net correction by returning true if it differs -- I do this for Snares which can be applied externally! Then you can handle the correction on the client in UCharacterMovementComponent::ServerMoveHandleClientError()

@thin stratus that is my understanding thus far

[thin stratus]

Hm yeah. Not sure about the ServerMove part

iirc the Server will set the Current NetworkMoveData 3 times or so.

Based on getting an old move, the current one or something else

can't recall

I think there might be a different spot, earlier to PerformMovement, where you could unpack it

But maybe it's fine

Would need to look at it again. I only have Mover in my head atm

@grand kestrel

[grand kestrel]

    // CallServerMovePacked ➜ ClientFillNetworkMoveData ➜ ServerMovePacked_ClientSend (client sends to server) ➜
    // ServerMovePacked_ServerReceive ➜ ServerMove_HandleMoveData ➜ ServerMove_PerformMovement ➜
    // MoveAutonomous (UpdateFromCompressedFlags)

Basically it leads into MoveAutonomous which immediately calls UpdateFromCompressedFlags()

Which means if you do it before calling Super::ServerMove_PerformMovement(), you're doing it one step before UpdateFromCompressedFlags()

[median igloo]

i hate how much unreal does for you

i cant make a 1 keyboard 2 player game its nearly impossible

because they handle allll the imput for you

i cant just add a camera and 2 players with programmed inputs becuase of that fucking input system

[thin stratus]

Is that just some random actor?

It will be pretty difficult to get actual physics working in Multiplayer fwiw.

Even if you get the events sorted, there is no proper prediction, reconcialiation and lag compensation. Especially not in Blueprints.

But that's besides the point and just the usual copy paste to warn people about Multiplayer + Physics.

Back in the day there were Facebook Groups I guess. Maybe there is something on Reddit.
Size-wise you can probably only really find this and the Forums

There is an Indie Server officially by Epic fwiw, but I doubt it really gives you more in terms of help

[lament flax]

of course you can

[thin stratus]

I mean, in general multiplayer terms you would want to call the impulse locally and then a server RPC with the same value

And the cube has to be owned by the client to allow server and client rpcs

Other than that not a lot can go wrong

Actor replicated and movement replication I gues

If you just want a throwable object with some bounciness you might be better off using an actor with a projectile movement component. That's how I usually fake throwable physic actors like balls

Works quite fine and comes with a bit of smoothing

[oak night]

Player numbers is not replicated.
So why is my "Run On Server" event updating a local variable thats on the client, and not the server?
I expect to update the varaible on the server, then send that to the owning client which takes the updated variable and uses it.
"No i cant use a replicated "player numbers" variable for this." 😉

I just dont get why, when i call this event from the client, it uses the clients variable on server, and not the server variable.

Can someone explain?

[fossil spoke]

Is CameraActor itself a Replicated Actor?

Is it owned by a PlayerController?

[oak night]

Its not a replicated actor

[fossil spoke]

Then it cannot send RPCs

If its not owned by a PlayerController then it cannot send an RPC to the Server.

[oak night]

But when i do this, the host does show a print string, and the client does show a print string on th owning client event as well.

[fossil spoke]

Well then CheckPlayers is being called by both of them.

[oak night]

Left host, right client.
Yellowish string is the run on server event.
Purple is the owning client event.

To me this looks like rpc is working?

If a actor is not replicated, does the local player not automatically "own" it?

[fossil spoke]

No

Depends on what its Owner is.

Which is determined by how you spawn it.

It would be helpful to see all of the calls to CheckPlayers and the implementation of RetrievePlayers as well.

[oak night]

Ok give me a sec

[fossil spoke]

If asking for help, its important to provide as much context as possible.

[oak night]

And thanks guys

[fossil spoke]

Otherwise we just waste time playing 20 questions

[oak night]

I understand, but know how much context is needed is difficult when experience on the subject is limited on my end 😉

Ok so im doing this from my default pawn called CameraActor.
Which is apparently replicated, i was unaware. (woopsy)

I have these events in it that i call on left mouse click.
I expect the server to add 1 to player numbers and send that back to the client.

Dont know what context should be needed more

[fossil spoke]

This is doing what I would expect it to do if the default value of the PlayerNumbers is 0 and you have just clicked once.

What are you expecting to happen?